My thoughts exactly: I used to be a pure reductionist ( its painful and it hurts ) until I learned the practical field of robotics and electronics. I started by building electronic robots. When I was programming the software for such robots, I noticed my platonic formulae didnt work. The robots behavior was totally different from the platonic formulae I imagined. Then I thought my platonic formulae were wrong and I should use more complex formulae that model the subtleties of robots. But I was wrong again. Platonic math is unrealistic and Utopian (lots of wild assumptions). Actuators never work as expected and sensors never do the same measurements for the same phenomena. Their behavior and measurements vary a lot through time and depend on environmental conditions. Robots are highly chaotic. That means if robots deviate just slightly at the beginning of the trajectory, the end result is even more different. Control theory didnt help that much. Control theory is just an attempt to overcome the constant deviations caused by the stochasticity and partially-observability of the environment. But control theory alone cannot learn the transition models of actuators and the transductions of sensors. And neither can platonic mathematics. Learning the transition models of actuators and the transductions of sensors are the job of probability theory and complex networks of patterns. But wait! Probability theory works fine for toy problems. On the other hand, robots are complex and generate complex geometries whose probabilistic inferences are computationally intractable. Thats why complex networks of patterns are the only way to proceed. Now, I still admire the power of platonic mathematics and I understand the need of using complex networks of patterns in artificial intelligence and machine learning. Im not a pure reductionist anymore. Im a hybrid. My mind understand both worlds: Reductionism (platonic math) and holism (networks). And there is a funny corollary here: Pure reductionists have never built any complex robot! -Juan
Posted on: Tue, 08 Jul 2014 00:55:39 +0000
Trending Topics
Recently Viewed Topics
© 2015